Re: LC-215 - Easy Add-ins (i18n Comment on XML Schema Last Call Working Draft)

Hello Martin,

Many thanks for these examples. I have had a look at them.

At 00/09/29 10:15 +0100, Martin Gudgin wrote:
>Hi Martin,
>
>Comments inline...

> > Hello Martin :-),
> >
> > Many thanks for following up on this. Unfortunately, your mail does
> > not give us enough information to decide whether we are satisfied
> > with the Schema WG decision. We need more detailed information,
> > and in particular some examples, to make such a decision.
> >
> > The part of our mail that is now registered as LC-215 contains
> > various various examples and proposals, and if you could work
> > out one of these, I guess that would be a good starting point.
>
>[MJG]
>To take the addition of xml:lang as an example;
>
>1.    Given the original schema document[1] and associated instance[2], it
>would be possible to produce a second schema document[3] that imports[1] and
>adds the xml:lang attribute to the type 'personName'. This would make[4] a
>valid instance document per schema document[3]
>
>2.    If changing the namespace of the instance is not an option then the
>second schema document[5] could use the redefine mechanism instead. The
>instance document at[2] would be valid per the schema document[5].

I think this shows how to add single attributes quite nicely.
I assume it is also possible to add attributeGroups this way,
could you confirm (we don't need the examples).
As discussed on a telephone conversation today with Michael,
that means that it is easy to add attributes to a whole bunch
of elements if these elements have a common base type (by
redefining the base type), but it will need more work to
add some attributes to a bunch of elements if they are
all defined independently. That looks quite reasonable,
it's difficult to expect more in the later case, I guess.


> > At 00/09/21 12:25 -0400, Martin Gudgin wrote:

> > >Among other issues, you raised the point registered as issue LC-215, Easy
> > >add-ins.
> >
> > The relationship of this to LC-216 should also be noted. Any ideas
> > about what the WG plans to do on that one?
>
>[MJG]
>We ( the XML Schema WG ) declined to adopt the proposal outlined in LC-216.
>I'm not sure who is the official respondent on that one, sorry.

We got the response from Michael, and are working on following up.


> > If you believe, that doesn't help us. If you know, and can show us
> > how, that would be helpful.
>
>[MJG]
>Hopefully example 1 above addresses this point.

It addresses quite some cases, thanks.

> > >2    The new 'redefine' mechanism[1] may make such changes easier in some
>or
> > >most cases.
> >
> > Do you mean "6.2.2 Including modified component definitions"
> >
>http://www.w3.org/XML/Group/xmlschema-current/structures/structures.html#mod
> > ify-schema
> >
> > It looks indeed like this might help, but we need more certainty than
> > that. As an example, assume a traditional HTML-like document type.
> > If this is well structured, it will somewhere have an 'inline'
> > equivalence class (if that term is still used in the current draft).
> > Can you show an example where:
> >
> > - An additional element is added to 'inline'?
> > - An additional attribute is added to 'inline'?
> >
> > Also, can you help us figure out whether the result of these additions
> > would be in the original namespace or a new one, or whether e.g.
> > only the new element/attribute would be in the new namespace?
>
>[MJG]
>Use of redefine is restricted to same namespace only so any changes made to
>declarations end up in the same namespace as they were in originally.

The main reason for LC-215 is to fix existing namespaces, so it's
not too bad, I guess.



>Example 2 above shows how to add an attribute, I hope to produce a
>substitution group ( new name for equivalence class ) example soon.

Thanks for that one, too.


> >3     A generic or fairly generic XSLT stylesheet could be written
> > >to automate the generation of types containing extra attributes, elements
>or
> > >sets of the same.
> >
> > It looks like that's not what we want. I.e. to take the above case,
> > if there is a concept of 'inline' in the schema, and the additions
> > are conceptually meant to go to 'inline', this should not have to
> > be done by adding things to the elements derived from inline,
> > neither by hand nor somehow automatically.
>
>[MJG]
>My point here was *not* that you could generate a new *instance* document
>using XSLT but that the new *schema* document could be generated using XSLT.

I think I got that. The proposal to use XSLT has two problems:
- It suggests that a lot of work is needed, maybe even for something
   that is conceptually just one change. Without 'redefine', there
   may be quite some such cases. Without it, there are much less.
- The script has to be run every time the original schema is changed.
   With declarative techniques, that's ideally not the case.


>Hopefully it can be seen from the structure of[1] and [5] that the latter
>could be generated from the former.

Now I understand. You mean to generate a schema that adds some things,
such as that it adds an xml:lang to everything of type mixed or string
or so.


>I hope to produce an example XSLT for
>this soon. Given that you will *have* to generate a new schema anyway
>getting a machine to generate it for you seems like a plus to me.

In some cases, that indeed helps.



> > >It would be helpful to us to know whether you are satisfied with the
> > >decision taken by the WG on this issue, or wish your dissent from the
> > >WG's decision to be recorded for consideration by the Director of
> > >the W3C.
> >
> > We want to know better how things would actually work, so that we can
> > understand whether we agree or disagree.
> >
> > I hope you can help us with that.
>
>[MJG]
>Hopefully this mail goes some way to doing that... I realise it doesn't
>address all your points but I wanted to get something out quickly in reply
>to you rather than leave you hanging. I'll try and address the other points
>over the next few days.

Thanks a lot. I'm rather optimistic that the i18n WG and IG can come
to the conclusion that LC-215 is dealt with sufficiently (if not
completely), but of course I cannot speak for the WG and IG.

Regards,   Martin.


>Cheers
>
>Martin Gudgin
>XML Schema Working Group
>
>[1] http://marting.develop.com/xsd/lc-215/ex1.xsd
>[2] http://marting.develop.com/xsd/lc-215/ex1.xml
>[3] http://marting.develop.com/xsd/lc-215/ex2.xsd
>[4] http://marting.develop.com/xsd/lc-215/ex2.xml
>[5] http://marting.develop.com/xsd/lc-215/ex3.xsd

Received on Monday, 16 October 2000 04:12:31 UTC